MAMEWorld Forums - The Loony Bin - Xtal compensation theory 101: Part 1

> > > For me, the clock was running a wee bit too fast, so...
> > > I told it that there were 500012 microseconds in half a second! Problem solved!
> >
> >
> > Problem solved at that specific temperature. Mess with your thermostat in the
> > slightest and you might have just made things worse.
> >
> > What you said isnt untrue, as long as ambeint is not a variable. I have some nice
> > temp vs frequency charts I can post later, if you would like.
>
> I'd like to see them.
> How bad is the problem, really? If it's not too bad, then (since the clock has a
> calendar) I could use a lookup table based on the month of the year. Assume 65
> degrees F in February, 75 degrees F in August, that sort of thing, as the clock is
> for indoor use.

It isn't that simple, not even in a really good looking crystal as shown.

Below is a pretty decent looking and fairly well compensated AT cut crystal, @ 10MHz. The plot you see is the frequency deviation from nominal (10Mhz), shown in ppm (Parts per million). In this case that equates to about .07ppm total deviation(70 parts per billion) over the temp range of -45C to 80C (+/-32ppb would be the appropriate tagging). That would make for one rocking timepiece, but you'd be paying $30 - $70 for just the oscillator.

To put that in perspective, your average uncompensated quartz clock oscillator is listed around 10ppm. That'll get you to +/- one second a DAY.

The crystal below is compensated as an "oven oscillator". They actually put the crystal in a little sealed container, and using a PID or similar control circuit keep the "oven" at a specific single temperature just above the highest rated use temperature. By this method they remove temperature as variable and can, as you put it, "tell the clock to run a little faster" (or slower). Even with aging, drift, and jitter, you'd end up with a very accurate clock. That is the nature of an oven oscillator, your error follows a smooth and predictable curve. The drawbacks here are expense, power required to run the oven, and space. You'd end up with a clock control circuit that would chew through a few AA batteries in a week, cost $100 itself without any of the clock parts, and take up an inordinate amount of real estate. Not your hot seller, IMO. Now, on to part 2 and why this won't work in other situations...

[ATTACHED IMAGE - CLICK FOR FULL SIZE]